Optical flow requires multiple strategies (but only one network)

Tal Schuster; Lior Wolf; David Gadot

doi:10.1109/CVPR.2017.732

Optical flow requires multiple strategies (but only one network)

Tal Schuster, Lior Wolf, David Gadot

School of Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

9 Scopus citations

Abstract

We show that the matching problem that underlies optical flow requires multiple strategies, depending on the amount of image motion and other factors. We then study the implications of this observation on training a deep neural network for representing image patches in the context of descriptor based optical flow. We propose a metric learning method, which selects suitable negative samples based on the nature of the true match. This type of training produces a network that displays multiple strategies depending on the input and leads to state of the art results on the KITTI 2012 and KITTI 2015 optical flow benchmarks.

Original language	English
Title of host publication	Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	6921-6930
Number of pages	10
ISBN (Electronic)	9781538604571
DOIs	https://doi.org/10.1109/CVPR.2017.732
State	Published - 6 Nov 2017
Event	30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017 - Honolulu, United States Duration: 21 Jul 2017 → 26 Jul 2017

Publication series

Name	Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017
Volume	2017-January

Conference

Conference	30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017
Country/Territory	United States
City	Honolulu
Period	21/07/17 → 26/07/17

Funding

Funders	Funder number
National Intelligence Service
Intel Collaboration Research Institute for Computational Intelligence

Access to Document

10.1109/CVPR.2017.732

Cite this

Schuster, T., Wolf, L., & Gadot, D. (2017). Optical flow requires multiple strategies (but only one network). In Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017 (pp. 6921-6930). (Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017; Vol. 2017-January). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CVPR.2017.732

@inproceedings{3de01d39cf7c4c2cb6287a9104185897,

title = "Optical flow requires multiple strategies (but only one network)",

abstract = "We show that the matching problem that underlies optical flow requires multiple strategies, depending on the amount of image motion and other factors. We then study the implications of this observation on training a deep neural network for representing image patches in the context of descriptor based optical flow. We propose a metric learning method, which selects suitable negative samples based on the nature of the true match. This type of training produces a network that displays multiple strategies depending on the input and leads to state of the art results on the KITTI 2012 and KITTI 2015 optical flow benchmarks.",

author = "Tal Schuster and Lior Wolf and David Gadot",

note = "Publisher Copyright: {\textcopyright} 2017 IEEE.; 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017 ; Conference date: 21-07-2017 Through 26-07-2017",

year = "2017",

month = nov,

day = "6",

doi = "10.1109/CVPR.2017.732",

language = "אנגלית",

series = "Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "6921--6930",

booktitle = "Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017",

address = "ארצות הברית",

}

Schuster, T, Wolf, L & Gadot, D 2017, Optical flow requires multiple strategies (but only one network). in Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017. Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, vol. 2017-January, Institute of Electrical and Electronics Engineers Inc., pp. 6921-6930, 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, Honolulu, United States, 21/07/17. https://doi.org/10.1109/CVPR.2017.732

Optical flow requires multiple strategies (but only one network). / Schuster, Tal; Wolf, Lior; Gadot, David.
Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017. Institute of Electrical and Electronics Engineers Inc., 2017. p. 6921-6930 (Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017; Vol. 2017-January).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Optical flow requires multiple strategies (but only one network)

AU - Schuster, Tal

AU - Wolf, Lior

AU - Gadot, David

PY - 2017/11/6

Y1 - 2017/11/6

N2 - We show that the matching problem that underlies optical flow requires multiple strategies, depending on the amount of image motion and other factors. We then study the implications of this observation on training a deep neural network for representing image patches in the context of descriptor based optical flow. We propose a metric learning method, which selects suitable negative samples based on the nature of the true match. This type of training produces a network that displays multiple strategies depending on the input and leads to state of the art results on the KITTI 2012 and KITTI 2015 optical flow benchmarks.

AB - We show that the matching problem that underlies optical flow requires multiple strategies, depending on the amount of image motion and other factors. We then study the implications of this observation on training a deep neural network for representing image patches in the context of descriptor based optical flow. We propose a metric learning method, which selects suitable negative samples based on the nature of the true match. This type of training produces a network that displays multiple strategies depending on the input and leads to state of the art results on the KITTI 2012 and KITTI 2015 optical flow benchmarks.

UR - http://www.scopus.com/inward/record.url?scp=85044512984&partnerID=8YFLogxK

U2 - 10.1109/CVPR.2017.732

DO - 10.1109/CVPR.2017.732

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:85044512984

T3 - Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017

SP - 6921

EP - 6930

BT - Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017

Y2 - 21 July 2017 through 26 July 2017

ER -

Schuster T, Wolf L, Gadot D. Optical flow requires multiple strategies (but only one network). In Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017. Institute of Electrical and Electronics Engineers Inc. 2017. p. 6921-6930. (Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017). doi: 10.1109/CVPR.2017.732

Optical flow requires multiple strategies (but only one network)

Abstract

Publication series

Conference

Funding

Access to Document

Other files and links

Fingerprint

Cite this